Following our demographic example, we might choose a demographic subgroup and have an algorithm suggest small changes or even full rewrites of existing text to make it score more highly for a given demographic. Recall that based upon the vast amount of clickthrough rate data we have, we can build models that can take a given text and predict its CTR (click-through rate) for a given demographic. Having this model we can then “run it in reverse” and back fill by starting with a given text and moving itgenerating modified contentthat our initial model tells us is more likely to received a high CTR from the target demographic.

Given a document, completely rewrite the document in a different format. E.g., translate a long movie review of Inception in the style of The New Yorker into a short movie review of the same movie in the style of Wired. Obviously this tends to be someone what easier going from long to short documents.

Stacks Image 277


Input Panel

Stacks Image 269

Function Select

Stacks Image 271

Parameters Select

Stacks Image 273

Output Panel

Stacks Image 275